whisper-small-hi-2400_500_101
This model is a fine-tuned version of saif-daoud/whisper-small-hi-2400_500_100 on the afrispeech-200 dataset. It achieves the following results on the evaluation set:
- Loss: 1.0512
- Wer: 0.3797
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-06
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 200
- training_steps: 900
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 6.1955 | 0.17 | 150 | 1.0763 | 0.6285 |
| 4.0503 | 0.33 | 300 | 1.0923 | 0.4813 |
| 3.2324 | 1.17 | 450 | 1.0762 | 0.4159 |
| 2.974 | 1.33 | 600 | 1.0650 | 0.3963 |
| 2.6571 | 2.17 | 750 | 1.0542 | 0.3888 |
| 2.7691 | 2.33 | 900 | 1.0512 | 0.3797 |
Framework versions
- Transformers 4.28.0.dev0
- Pytorch 1.13.1+cu116
- Datasets 2.10.1
- Tokenizers 0.13.2